The OGTT With Insulin: The Test Most PCOS Work-Ups Skip

Why does an OGTT catch what HbA1c and fasting glucose miss in PCOS?

A 2-hour oral glucose tolerance test measures how blood glucose actually behaves under a real challenge, while HbA1c and fasting glucose only sample two narrow moments — a three-month average and a single fasting number — that PCOS physiology is unusually good at hiding behind. A 2024 systematic review and meta-analysis of 2,628 women with PCOS tested HbA1c and fasting glucose against the OGTT as the reference standard and found that an HbA1c of 6.5% or higher — the standard type 2 diabetes cutoff — had a pooled sensitivity of only 50% in this population, meaning it missed half of the women the OGTT identified as having type 2 diabetes. Fasting plasma glucose of 7.0 mmol/L or higher performed only slightly better, at 58% sensitivity. Both tests were highly specific — they rarely flagged someone who did not have diabetes — but specificity does not help the person whose result comes back falsely reassuring.

An earlier, PCOS-specific study makes the same point at the pre-diabetes end of the spectrum. A 2007 study of 105 women with PCOS tested the fasting glucose cutoffs that Canadian and American diabetes guidelines use to decide who needs an OGTT at all. The standard cutoff of 5.6 mmol/L had 48% sensitivity for detecting abnormal glucose tolerance in this group — using it as a gatekeeper would miss 52% of the women who actually had impaired glucose tolerance on the OGTT. Even the study’s own recalculated, PCOS-specific optimal cutoff of 5.0 mmol/L still missed 24%. The study’s conclusion was blunt: fasting-glucose gatekeeping does not work in PCOS, and every woman with PCOS should have the OGTT itself rather than a fasting glucose used to decide whether she “qualifies” for one.

Who actually recommends the OGTT, and why insulin gets added to it?

The Androgen Excess Society’s 2007 position statement recommends a 2-hour, 75-gram OGTT for every woman diagnosed with PCOS, regardless of body mass index, with rescreening at least every 2 years if the result is normal and annually if it shows impaired glucose tolerance. This sits alongside, not instead of, the rest of the standard PCOS blood panel drawn at the same visit. Adding a fasting or paired insulin measurement to the same draw does not diagnose diabetes any more precisely than glucose alone does, but it is what turns a single OGTT into raw material for an insulin-sensitivity calculation — HOMA-IR, QUICKI, or the Matsuda index — since every one of those indices needs at least a fasting insulin value, and the Matsuda index needs insulin at every OGTT timepoint, not just fasting.

How do HOMA-IR, QUICKI and the Matsuda index actually compare?

HOMA-IR, QUICKI and the Matsuda index all estimate insulin resistance from the same OGTT blood draw, but they need different amounts of that draw and were validated against the gold-standard euglycemic clamp with different accuracy. HOMA-IR needs only a single fasting glucose and fasting insulin value, plugged into the formula (fasting glucose in mmol/L × fasting insulin in µU/mL) ÷ 22.5. In its original 1985 validation, HOMA-IR correlated with clamp-measured insulin resistance at Rs = 0.88 — a strong population-level correlation — but the same paper reported a coefficient of variation around 31% for the individual estimate, meaning two blood draws from the same person on different days can produce meaningfully different HOMA-IR numbers even when nothing about their physiology changed.

QUICKI uses the same two fasting values as HOMA-IR, run through a different formula — 1 ÷ [log(fasting insulin in µU/mL) + log(fasting glucose in mg/dL)] — and its original 2000 validation reported a somewhat stronger correlation with the clamp, r = 0.78, in a mixed group of 56 non-obese, obese and type 2 diabetic subjects. Because QUICKI is built from the same two numbers as HOMA-IR, run through a mathematically related formula, the two rarely disagree about direction — they mostly just report the same signal on different scales.

The Matsuda index is the only one of the three that requires the full OGTT rather than a fasting sample alone: glucose and insulin at fasting and again during the test, most commonly at 30, 60, 90 and 120 minutes. Its original 1999 validation in 153 subjects — spanning normal glucose tolerance, impaired glucose tolerance and type 2 diabetes — found it correlated with the clamp at r = 0.73, and the paper specifically showed that a simple fasting glucose-to-insulin ratio, without the OGTT data, did not correlate with clamp-measured insulin sensitivity at all (r = -0.02). That is the paper’s central point: fasting values alone leave out how the body handles a real glucose load, which the Matsuda index captures and HOMA-IR and QUICKI do not.

Why isn’t there one universal HOMA-IR cutoff?

No HOMA-IR number applies across every lab, population and insulin assay, because the number is built from an insulin measurement that is not standardized the way glucose is. A 2004 review of HOMA modeling, published by one of the index’s original authors, states plainly that HOMA-IR had already appeared in more than 500 publications by that point, and cautions that the raw insulin assay underlying it varies enough between laboratories and manufacturers that a cutoff validated in one population, on one assay, does not transfer cleanly to another. That is why you will see different HOMA-IR thresholds — commonly somewhere between 2.0 and 2.9 — cited as “the” cutoff for insulin resistance across different papers and calculators: each was derived from a different reference population, on a different insulin assay, and none of them is more universally correct than the others. What a HOMA-IR result actually means once you have it in hand walks through reading your own number against your lab’s specific reference range, rather than a threshold copied from a different study.

Do most PCOS work-ups actually order fasting insulin?

Many routine PCOS work-ups stop at fasting glucose or HbA1c and never add a fasting insulin value, because insulin assays cost more, are not run by every lab, and are not part of the standard metabolic panel a primary-care visit typically orders. Without a fasting insulin value, none of HOMA-IR, QUICKI or the Matsuda index can be calculated at all — the OGTT-with-glucose-only version of the test, which many clinics do run, tells you whether glucose tolerance is abnormal but tells you nothing about the insulin dynamics behind it. If insulin resistance itself, rather than glucose tolerance alone, is the question you want answered, you need to specifically ask that a fasting — and ideally paired — insulin value be drawn alongside the glucose, as one part of a full PCOS diagnostic work-up rather than a stand-alone request. What to actually do about a result that comes back elevated is a separate question, covered in what moves HOMA-IR once it is high.

Note: in May 2026, PCOS was renamed polyendocrine metabolic ovarian syndrome, or PMOS, by a global consensus of more than 50 organisations. The tests and thresholds in this article are unchanged by the rename; this article uses PCOS because that is still the term most readers search.

Reading this by phenotype

The insulin-resistant phenotype is where the OGTT with insulin does the most diagnostic work: fasting glucose and HbA1c are the most likely to look reassuring here despite real glucose intolerance, and a fasting insulin drawn alongside the OGTT frequently returns clearly elevated even before glucose itself moves out of range — the same phenotype in which free testosterone tends to run ahead of total testosterone for the same low-SHBG reason. In the lean, hyperandrogenic phenotype, glucose and insulin values are usually unremarkable, and the OGTT is run mainly to rule out the minority of lean cases that do carry glucose intolerance rather than to confirm an expected finding. In the ovulatory-but-hyperandrogenic and non-hyperandrogenic phenotypes, metabolic testing follows general population risk rather than a PCOS-specific pattern, since neither phenotype is defined by an insulin or glucose abnormality.

Who this doesn’t work for

The OGTT is a poor choice for someone who cannot tolerate the 75-gram glucose drink itself — a meaningful minority of people feel nauseated or unwell during the two-hour wait, and a small number cannot complete it. HOMA-IR, QUICKI and the Matsuda index all assume a person is not on a medication that directly alters glucose or insulin at the time of the draw; metformin, insulin therapy and, to a lesser extent, hormonal contraception all distort the numbers these formulas were validated without. And none of these three indices, on their own, diagnoses PCOS — they describe insulin sensitivity, which is common but not universal across the four Rotterdam phenotypes, and a lean, insulin-sensitive person can carry a PCOS diagnosis with entirely unremarkable results on every index in this article.